Lab on a Chip

Publ

ishe

d on

30

Oct

ober

201

4. D

ownl

oade

d by

Law

renc

e B

erke

ley

Nat

iona

l Lab

orat

ory

on 3

0/10

/201

4 13

:57:

29.

PAPER View Article OnlineView Journal

This journal is © The Royal Society of Chemistry 2014

a Sandia National Laboratories, 7011 East Ave, Livermore, CA, USA.

E-mail: aksingh@sandia.gov; Fax: +1 925 294 3020; Tel: +1 925 294 1260b Joint Bioenergy Institute (JBEI), 5855 Hollis St., Emeryville, CA, USAc Lawrence Berkeley National Laboratories (LBNL), Building 64R0121, 1 Cyclotron Road,

Berkeley, CA 94720, USAdDepartment of Bioengineering, University of California, Berkeley, CA, USA

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c4lc00794h

Cite this: DOI: 10.1039/c4lc00794h

Received 8th July 2014,Accepted 6th October 2014

DOI: 10.1039/c4lc00794h

www.rsc.org/loc

A droplet-to-digital (D2D) microfluidic device forsingle cell assays†

Steve C. C. Shih,ab Philip C. Gach,ab Jess Sustarich,ab Blake A. Simmons,ab

Paul D. Adams,bcd Seema Singhab and Anup K. Singh*ab

We have developed a new hybrid droplet-to-digital microfluidic platform (D2D) that integrates droplet-in-

channel microfluidics with digital microfluidics (DMF) for performing multi-step assays. This D2D platform

combines the strengths of the two formats—droplets-in-channel for facile generation of droplets

containing single cells, and DMF for on-demand manipulation of droplets including control of different

droplet volumes (pL–μL), creation of a dilution series of ionic liquid (IL), and parallel single cell culturing and

analysis for IL toxicity screening. This D2D device also allows for automated analysis that includes a

feedback-controlled system for merging and splitting of droplets to add reagents, an integrated Peltier

element for parallel cell culture at optimum temperature, and an impedance sensing mechanism to control

the flow rate for droplet generation and preventing droplet evaporation. Droplet-in-channel is well-suited

for encapsulation of single cells as it allows the careful manipulation of flow rates of aqueous phase

containing cells and oil to optimize encapsulation. Once single cell containing droplets are generated, they

are transferred to a DMF chip via a capillary where they are merged with droplets containing IL and

cultured at 30 °C. The DMF chip, in addition to permitting cell culture and reagent (ionic liquid/salt)

addition, also allows recovery of individual droplets for off-chip analysis such as further culturing and

measurement of ethanol production. The D2D chip was used to evaluate the effect of IL/salt type

(four types: NaOAc, NaCl, [C2mim] [OAc], [C2mim] [Cl]) and concentration (four concentrations: 0, 37.5, 75, 150mM)

on the growth kinetics and ethanol production of yeast and as expected, increasing IL concentration led

to lower biomass and ethanol production. Specifically, [C2mim] [OAc] had inhibitory effects on yeast

growth at concentrations 75 and 150 mM and significantly reduced their ethanol production compared to

cells grown in other ILs/salts. The growth curve trends obtained by D2D matched conventional yeast

culturing in microtiter wells, validating the D2D platform. We believe that our approach represents a

generic platform for multi-step biochemical assays such as drug screening, digital PCR, enzyme assays,

immunoassays and cell-based assays.

Introduction

Single cell analysis has become a valuable means to analyzecells since it is able to decipher individual cells within aheterogeneous population. The single cell approach can assistour understanding of complex biological systems on the cellularlevel that can not be obtained through population averages.This requires meticulous investigations on the growth of singlecells in response to a variety of chemical and biological cues.

One area of interest is screening single cells for effectivebiofuel production and microbial toxicity to a variety of chemicalsfrom upstream pretreatment processing of lignocellulosicbiomass.

Recently, lignocellulosic biofuels have attracted a lot ofattention as they provide an attractive alternative to fossilfuels as they are derived from non-food crops and have a lowerimpact on the environment.1,2 The key steps in cellulosic bio-fuel production involve the thermochemical pretreatment ofbiomass, followed with enzymatic saccharification of the pre-treated biomass, and fermentation of the resulting sugars intoethanol or other biofuels in a host such as yeast. A promisingpretreatment technology is ionic liquid (IL) pretreatment3–5

because of their ability to efficiently dissolve and fractionatebiomass and produce higher yield of lignin-free cellulosefrom both grasses and woody biomass compared to othermethods.6 The pretreated cellulose is further broken down

Lab Chip

Lab on a ChipPaper

Publ

ishe

d on

30

Oct

ober

201

4. D

ownl

oade

d by

Law

renc

e B

erke

ley

Nat

iona

l Lab

orat

ory

on 3

0/10

/201

4 13

:57:

29.

View Article Online

into simple sugars for downstream fermentation by cellu-lases (e.g. β-glucosidases).7,8 At high concentrations, some ILsused for pretreatment are toxic to most enzymes and organ-isms used for saccharification and fermentation, respectively.However, complete removal of IL from pretreated biomassand sugar is prohibitively expensive. In addition, a consoli-dated one-pot process with combined pretreatment and sac-charification unit operations or consolidated pretreatment–saccharification–fermentation unit operations are highlydesirable. To eliminate the burden of complete removal of ILfrom sugar generated from IL pretreatment and to enable theefforts around one-pot consolidated processes, large numbersof enzyme variants and mutant strains have been generatedthat need to be screened for tolerance to IL, and selection ofpromising candidates for further process scale up. Significantefforts have been made to analyze ILs and to determine theireffect on hydrolysis rates of pretreated lignocelluloses,9,10 andexamine their toxicity in fungi and in bacteria.11 Currently,the screening is done using large populations of cells withmicrotiter plates and cuvettes.9–12 However, the microtiterplate based screening has a number of drawbacks – it is slow,expensive as it uses a large amount of reagents, and suffersfrom poor quantification and reproducibility. In addition,in laboratories that lack robot dispensers and aspirators, thisprocess requires a significant amount of manual labor andprocedural optimization. Finally, planktonic cells such as yeastand E. coli exhibit erratic culture behavior (e.g. variations inbiomass growth), which result in unusual growth patterns.13,14

Hence, an alternative to understand the effects of IL onbiofuel-producing cell strains in populations is to isolate andanalyze single cells for toxicity towards a specific IL in variousconcentrations.

Microfluidic platforms offer a potential solution to over-come many of these drawbacks – they can drastically reducethe number of pipetting steps required, lower the cost ofreagents by at least an order of magnitude, reduce the timerequired for screening, and potentially improve the through-put. Droplet-in-channel microfluidics has emerged as a verypromising technology where pL–nL monodisperse dropletscan be used as an alternative to microtiter wells to conducthigh-throughput assays.15,16 This is a two-phase system whereaqueous droplets are formed with a surrounding oil phase(there are also multiple phase systems17–19). The droplet-in-channel format has been used for many applications includ-ing culturing and analyzing of single cells,20–22 multistepenzyme assays,23–25 and clinical diagnostics.26,27 A challengewith this format is the serial addition of reagents – akin topipetting. Many different schemes have been proposed to addone reagent at a time to the aqueous droplet: droplet fusion,28,29

electrocoalescence,30 picoinjection,31,32 and specialized micro-fabricated structures to induce droplet merging.33,34 These areinnovative methods that require exquisite control over flowrates, timing, fluidic resistance, rate of droplet generation, andpressure fluctuation. Here, we present an alternative form ofdroplet microfluidics which has been shown to be an elegantplatform for reagent addition via droplet merging – digital

Lab Chip

microfluidics (DMF).35,36 Digital microfluidics is well-suitedfor adding reagents in parallel and reagents can be mixed ondemand without the requirement of optimal and precise flowrates. A DMF device typically comprises of an open array ofelectrodes covered with a dielectric and a hydrophobic layer.To move droplets, electric potentials are applied to the electrodesuch that electrostatic forces (i.e. qE) are generated which allowdroplets to perform a number of different operations (e.g. move,merge, mix, split, and dispense).36 The low reagent consump-tion and facile connection to analytical instruments has madeDMF a popular platform for cell culture and assays.37–42 Typi-cally, a density of cells is cultured in a droplet on the DMFplatform and contains a surfactant, Pluronics,43,44 to mini-mize bio-fouling on the digital microfluidic surface. AlthoughPluronics has no adverse effects in the gene expression ofmammalian cells45 (and no change in their proliferation42),the inhomogeneous nature of the Pluronics is not suitablefor post-processing techniques such as mass spectrometry(MS) and/or HPLC.46

In this report, we discuss a novel proof-of-principle architec-ture that combines the strengths of the two droplet microfluidicplatforms (which we call droplet-to-digital microfluidics – D2Dmicrofluidics) while overcoming the weaknesses of the indi-vidual devices. This device allows us to analyze cells withoutthe use of Pluronics (e.g. F68 and F127) which permits com-patibility with post-processing techniques such as HPLC andMS. Furthermore, it may also enhance the lifetime of deviceuse since Pluronics is not a permanent solution to preventbiofouling.43,44 The most important advantage is the deviceconfiguration, which enables droplet generation by droplets-in-channels with parallel addition of reagents (on demanddispensing from reservoirs and splitting of droplets) andsingle cell analysis by digital microfluidics. Furthermore, thisnew technique allows manipulation of wide range of volumes(pL volumes for droplet-in-channels and pL–μL volumes forDMF) and is easily adaptable for automation. Here, we haveapplied our D2D device to analyzing ILs and determiningtheir effects on downstream processes related to biofuel pro-duction (e.g. microbial growth and ethanol production). Wepropose that this system can be applied to a wide range ofapplications that requires screening of single cells that involvesaddition of multiple reagents with a variety of volumes, auto-mation, and post-processing analysis.

Materials and methodsReagents and materials

Unless otherwise specified, general-use reagents were purchasedfrom Sigma Aldrich. Sacchromyces cervisiae strain BY4742was obtained from the JBEI Registry (https://registry.jbei.org).SYTO-9, a stain for nucleic acid in yeast was purchased fromLife Technologies (Grand Island, NY).

D2D microfluidic device fabrication

Droplet-to-digital (D2D) microfluidic devices were fabricatedin the University of California Biomolecular Nanotechnology

This journal is © The Royal Society of Chemistry 2014

Lab on a Chip Paper

Publ

ishe

d on

30

Oct

ober

201

4. D

ownl

oade

d by

Law

renc

e B

erke

ley

Nat

iona

l Lab

orat

ory

on 3

0/10

/201

4 13

:57:

29.

View Article Online

Center (UC Berkeley BNC) fabrication facility, using a trans-parent photomask printed at Fineline Imaging (ColoradoSprings, CO). Fabrication reagents and supplies includedSU-8-5, SU-8-2025, SU-8-2075, S-1811 and SU-8 Developer fromMicrochem (Newton, MA), gold- and chromium-coated glassslides from Telic (Valencia, CA), indium tin oxide (ITO) coatedglass slides (Delta Technologies, Stillwater, MN), Aquapelfrom TCP Global (San Diego, CA), MF-321 positive photoresistdeveloper from Rohm and Haas (Marlborough, MA), StandardKI/I2 gold etchant from Sigma, CR-4 chromium etchant wasfrom OM Group (Cleveland, OH), and AZ-300T photoresiststripper from AZ Electronic Materials (Somerville, NJ).

To fabricate a DMF device, it involved three steps:1) electrode patterning on the bottom plate, 2) dielectricdeposition layer, and 3) channel/spacer features. The firststep involving DMF fabrication bearing patterned electrodesand contact pads were formed by photolithography andetching as described previously41 with an exposure time of5 s (40mW cm−2) using an OAI Series 200 Aligner (San Jose,CA USA). For the second step, these devices were plasmatreated under 20% O2 and RF power of 20 W for 3 minand were then coated with a 7 μm layer dielectric usingSU-8-5 following Microchem's instructions for spin speedand bake times. For the channel/spacer features (step 3),the devices were then plasma treated (using the same con-ditions as above) and coated with a layer of SU-8-2075 topattern channels and a spacer with a height of 140 μm.Spin speeds, soft and hard bake times, and developmenttimes were following Microchem's instructions. After devel-opment, these were rinsed and dried with isopropanol anddiH2O and were hard-baked for 15 min at 200 °C. To createa hydrophobic layer, DMF devices and ITO top-plates werecoated with 0.2 μm filtered Aquapel. After 15 min, theAquapel was removed with a kimwipe and rinsed with diH2O.This was then dried with N2 and left at room temperaturefor 1 h.

The droplet microfluidic devices were fabricated usingsoft lithography techniques. Briefly, we first fabricated themicrofluidic channels using SU-8-2025 photoresist on a4" silicon wafer (i.e. a master mold). Following Microchem'sguidelines, the structure of the master has a height of 140 μm,generating the same dimension height in the PDMS mold. Themaster was silanized with trichloroIJ1H,1H,2H,2H-perfluorooctyl)silane (97%, Aldrich) overnight. PDMS (Sylgard 184, Dow Corning,Midland, MI) mixed at a ratio of 10 to 1 (base to curing agent)was poured onto the master, placed in a vacuum to removebubbles for 12 h, cured at 80 °C for 1 h, and peeled away. Theinlet and outlet holes were punched into the layer of channelsusing a Harris Uni-Core with a tip diameter of 0.75 mm.We exposed the PDMS and a glass substrate to oxygen plasma(Harrick Plasma, Ithaca, NY) for 3 min at max RF power andbonded the PDMS and substrate in contact to form a perma-nent seal.

To assemble the capillary interface (which was pur-chased from IDEX – item no. 1572, PEEK tubing 360 μm O.D,150 μm I.D. × 5 ft), one end of the capillary was inserted into

This journal is © The Royal Society of Chemistry 2014

the outlet of the droplet microfluidic device and the other endwas inserted into the channel on the DMF device (a channeldirected to the 300 μm electrodes). After insertion, the capil-lary was adhered to the device using 40 μm adhesive plate seal(PlateMax Ultraclear Sealing Film, Axygen, Union City, CA USA)and epoxy by UV curing for 1–2 min (Fig. 1a). An explodedview is shown in ESI† (Fig. S1).

D2D microfluidic automated control

The droplets on the digital microfluidic device were con-trolled by an in-house C++ program which was used to con-trol the application of driving potentials using an automatedfeedback control system for high-fidelity droplet movement.The feedback system is described elsewhere41,47,48 and wasused to monitor droplet movement on the device and dropletevaporation. Briefly, to move a droplet on a destinationelectrode, a 200 ms pulse of driving potential is applied to thedestination electrode relative to the top-plate electrode. Theoutput from the feedback system is connected to an analoginput of an RBBB Arduino microcontroller (Modern Device,Providence, RI), which controls the control board containingsolid-state switches (responsible for applying potential to theDMF device). The Arduino microcontroller was also used tocontrol the Peltier cooler, which is used to maintain constantsurface temperature on the DMF during cell culture. TheneMESYS syringe pumps was controlled by an in-house C++program that enables control of the flow rate of the liquids/droplets in the droplet generator and the oil injection betweenthe top and bottom layer of the DMF device to prevent evapo-ration of droplets on the DMF during cell culture. A figureshowing the automated feedback system with the signal flowis shown in Fig. 3. The automation code includes Arduinoand C++ code to run the microfluidic experiments (see ESI†).

Macroscale ionic liquid screening experiments

Macroscale growth assays were performed using the Saccha-romyces cerevisiae strain BY4742. A cryopreserved stock ofthe strain was streaked on agar plates of SD medium andsupplemented with 20 g L−1 of glucose and incubated at30 °C. All media components were of cell culture or mole-cular biology grade. For well-plate screening assays, 5 mLstarter cultures were grown at 30 °C to saturation in SDmedium supplemented with 20 g L−1 glucose in 30 mL testtubes with shaking at 200 rpm for 16–20 h. The opticaldensity of the starter cultures was determined at 600 nm(OD 600) with the Tecan Infinite F200 Pro (Tecan Ltd,San Jose, CA) well-plate reader. The starter cultures were dilutedto a final OD of 0.05 (measured with a spectrophotometer usinga 1 cm cuvette) with SD medium, supplemented with either20 g L−1 of glucose spiked with 1-ethyl-3-methylimidazoliumacetate (abbreviated as [C2mim] [OAc]), 1-ethyl-3-methylimidazoliumchloride ([C2mim] [Cl]), sodium chloride (NaCl), or sodiumacetate (NaOAc) to give final concentrations of 0 (control),37.5, 75, and 150 mM. Cultures were distributed into 24-well

Lab Chip

Fig. 1 Droplet to Digital (D2D) microfluidic device. (a) Image of device with capillary connecting the droplet-in-channel to DMF device. Coloreddye is used to outline the channel droplet device and to show droplets in reservoirs of digital microfluidic device. (b) Schematic of the D2D device.Fluid is pressure-driven by syringe pumps into the droplet generator. Droplets from the generator are driven to the DMF device (via capillary) andare actuated by high voltage (HV) signals. The DMF device is connected to two 40-pin connectors which guide HV signals to the contact pads ofthe DMF device. A Peltier element (to control temperature) is situated below the four cell culture regions on the DMF device. (c) Schematic of thedroplet-in-channel device used to generate ~20 nL droplets containing a single cell. In (d), the digital microfluidic design includes four cultureregions with L-shaped electrodes, four reservoir electrodes (used for ionic liquid, cell media, and waste), a channel layer (shown in red) that is usedfor 1) fluid delivery, 2) air bubble removal, and 3) interfacing the capillary from the droplet device to the digital microfluidic device. A dielectric layeris not shown for clarity. An exploded view of the DMF device is shown in ESI† Fig. S1.

Lab on a ChipPaper

Publ

ishe

d on

30

Oct

ober

201

4. D

ownl

oade

d by

Law

renc

e B

erke

ley

Nat

iona

l Lab

orat

ory

on 3

0/10

/201

4 13

:57:

29.

View Article Online

plates (800 μL per well). Plates were incubated at 30 °C withshaking for 24 h in a Tecan Infinite F200 Pro plate reader andOD600 was acquired every 20 min. To generate growth curves,the measured OD was corrected by subtracting the blank OD(i.e. only SD media) for each ionic liquid concentration andwere plotted in the natural log scale. All experiments wererepeated at least three times.

To quantitatively determine the impact of the ionic liquidson yeast growth, the growth kinetics were calculated usingthe following equation:

d ln

d

f

i

NNt

(1)

where μ is the growth rate, N is the O.D. of the cells at aninitial and final time and t is the time. The O.D. valueswere taken at times in the exponential region (i.e. the linearregion of the curve). A table of values to calculate the growthrate for a specific IL and concentration is shown in the ESI†(Tables S1 and S2). These growth rate values were tested for

Lab Chip

significance using a one-way ANOVA test with a post hocTukey analysis at a P-value < 0.05.

Microfluidic ionic liquid screening experiments

To ensure monodisperse droplet generation and preventionof merging of droplets in the generator, channels were coatedwith Picoglide-1™ (0.5% w/w FC-40) (Dolomite Microfluidics,Charlestown, MA) using a neMESYS (Cetoni, Korbussen,Germany) syringe pump with a flow rate of 0.2 μL s−1 for5 min and further incubated with the Picoglide for 1 h. Thechannels were then rinsed at the same flow rate with onlyHFE 7500 for 5 min and then dried in room temperaturefor 30 min.

To prepare cells for microfluidic analysis, 5 mL startercultures were grown at 30 °C to saturation in SD mediumsupplemented with 20 g L−1 glucose in 30 mL test tubes withshaking at 200 rpm for 16–20 h. After incubation, traces ofcell debris and clumping can occur (which can cause clog-ging in the microfluidic device) and therefore the cells weretransferred to fresh media. To minimize clumping of cells,

This journal is © The Royal Society of Chemistry 2014

Lab on a Chip Paper

Publ

ishe

d on

30

Oct

ober

201

4. D

ownl

oade

d by

Law

renc

e B

erke

ley

Nat

iona

l Lab

orat

ory

on 3

0/10

/201

4 13

:57:

29.

View Article Online

the culture was sonicated in a heated water bath (30 °C) for5 min. 80 μL of cells were then diluted in 1 mL of freshmedia and then transferred to a 2.5 mL High-Precision GlassSyringe with Tubing connector 1/4-28 UNF thread (Cetoni)containing 500 μL of HFE 7500. This dilution was performedsuch that the majority of droplets generated contained nomore than one cell (i.e. we are using Poisson's statistics).15

Although this protocol results in the formation of someempty droplets, these droplets were manually actuated withdroplets with a cell. Droplets containing two or more cells(<10% of the droplets) were analyzed as a droplet containingsingle cell. To generate the droplets containing a single cell,the oil (i.e. HFE 7500 with 0.5% v/v Picosurf) flow ratewas set to 0.1 μL s−1 and then it was set to 0.05 μL s−1 whenthe oil reaches the T-junction. At the same time, the aqueous(i.e. cells in media) flow rate was set to 0.005 μL s−1 whichgenerated ~20 nL droplets containing a single cell. To imagethe cells inside the channel, cells were stained with SYTO-9following the manufacturer's instructions. We used the sameprepared cells for replicates and re-prepared fresh cells andmedia for other IL/salt conditions.

To prepare the digital microfluidic device for single cellscreening, the capillary was injected with HFE 7500 using a3 mL syringe before connecting it to the droplet microfluidicdevice. This ensures there are no air bubbles trapped in thecapillary during operation. To inject liquids (cell media and300 mM ionic liquid/salt) into the reservoirs, we used twotechniques. The first technique consisted of pipetting 3.0 μLliquids directly onto the reservoirs, filling the surroundingarea with biocompatible oil (i.e. HFE 7500), and then placingthe ITO top-plate. The second technique consisted of placinga capillary into the channel that is connected to the reservoir(see red channels in Fig. 1d) and injecting the liquid intoreservoirs with the syringe pump. The advantage of usingtechnique two is the continuous filling of the reservoirs asthe liquid in the reservoir is depleted using our automatedcontrol with feedback. The filling of the reservoir with liquidis performed with the ITO top plate and in biocompatible oil.After injection of the liquids, a serial dilution is conducted tocreate four concentrations: 0 mM (control), 37.5 mM, 75 mM,and 150 mM of ionic liquid using the automated system bygenerating a script using an in-house C++ program.

To generate growth curves, images of the cells were cap-tured at different time intervals (0, 2, 6, and 24 h) using aninverted Olympus IX71 microscope connected to a scientificgrade CCD camera (iXon EMCCD, Andor Technology, SouthWindsor, CT) at a 40×magnification. The cells from the imageswere counted manually or using Image J and these values wereplotted vs. time. Growth kinetics were calculated using eqn (1)(above) and were tested for significance using a one-wayANOVA test and a Tukey's post hoc test using a P-value < 0.05.

Ethanol production in yeast

To quantify ethanol production in yeast, the cultures (after 24 h)on the D2D device were pipetted into 0.6 mL centrifuge tubes

This journal is © The Royal Society of Chemistry 2014

containing 0.3 mL of fresh media to ensure anaerobic growth.The tubes were sealed with parafilm and incubated at 30 °Cwith shaking at 200 rpm. After 24 h, the cultures were filteredusing a 0.2 μm membrane filter to collect the supernatant(and to remove yeast cells) before HPLC analysis. The quanti-fication of ethanol was conducted using an Agilent 1200HPLC equipped with a refractive index detector (Agilent,Santa Clara CA). Separations were achieved using an AminexHXP-87H (300 × 7.8 mm, 9 μm) analytical column (Bio-Rad,Hercules, CA USA) with a Micro-Guard Cation H guard col-umn (30 × 4.6 mm) at 50 °C. A mobile phase of 4 mM H2SO4

at 0.6 mL min−1 was used. A calibration curve was generatedthrough the use of external standards.

Results and discussionDroplet-to-digital (D2D) microfluidic device and operation

The goal of this work was to develop a microfluidic system toscreen and to analyze cells that are tolerant to different typesand concentrations of IL. Our microfluidic system joins asmall group of studies that have used a combination (“hybrid”)of microfluidic platforms. Although in these studies these“hybrid” systems are presented as digital-to-channel systems(which is different from our droplet-to-digital system), they dopresent interesting innovations that helped guide our work.49,50

Abdelgawad et al.49 presents a preliminary single-plate DMFconfiguration with a PDMS channel. However, it suffers a keylimitation in which single plate DMF devices are not capableof dispensing droplets from reservoirs – a requirement if weare to create a library of samples with different concentrations.Watson et al. presented a second generation hybrid micro-fluidic device that uses a digital microfluidic device formed ona top substrate which is mated to a network of microchannels.Here, the fabrication is more complex (e.g. drilling holesfor transferring reagents) and requires precise alignmentprocedures. Therefore, in this work, we chose instead to createtwo devices and mate them by using a capillary interface with-out specialized fabrication or precise alignment (as shown bysome groups,51–54 a capillary can be easily interfaced to adigital microfluidic device). Here, we describe the first imple-mentation of a droplet-to-digital microfluidic device that iscapable of parallel addition of multiple reagents on demandand single cell analysis.

To accurately compare the effect of various parameters oncell growth and ethanol production, it is important to startwith a single cells in each droplet. In designing the D2D sys-tem shown in Fig. 1, we evaluated a number of alternativestrategies to obtain a single cell in a droplet, includingpipetting a low density of cells into a reservoir on the DMFand use actuation to dispense a single cell from the reservoir.Dispensing on DMF can be accomplished by breaking a largerdroplet into smaller droplets with reproducible volumes.55

We attempted to generate a droplet with a single cell bystarting with a larger droplet (reservoir) containing a lowdensity of cells. However, cells in the reservoir tend to moveaway from dispensing electrode during actuation making it

Lab Chip

Lab on a ChipPaper

Publ

ishe

d on

30

Oct

ober

201

4. D

ownl

oade

d by

Law

renc

e B

erke

ley

Nat

iona

l Lab

orat

ory

on 3

0/10

/201

4 13

:57:

29.

View Article Online

hard to dispense them. This behavior was attributed to thelow-conductance of the cell media, which results in movingcells away from the high field.56 Furthermore, dispensinga volume from a larger droplet is a random process on thedigital microfluidic device and it is very difficult to obtainconsecutive successful dispensed droplets containing onlya single cell and to predict when a dispensed droplet has asingle cell. On the other hand, encapsulation of single cells iseasily accomplished in droplet-in-channel microfluidics bycareful balancing of aqueous flow (containing cells) and oilflow and has been demonstrated by many groups.57,58 Hence,we used a droplet generator to create droplets encapsulating asingle cell (with some droplets generated with no cells or twoor more cells) (Fig. 1c). This device made out of PDMS wasstable over the course of many droplet generations, and couldbe used months later without loss of operation.

Lab Chip

Fig. 2 Device operation of the D2D microfluidic device. (a) A series of im(SYTO-9) was used to image the yeast cell. Scale bar is 200 μm. (b) DropletFrame 2 shows a droplet encapsulating a single cell and actuated on DMis then mixed with a larger droplet (100 nL) (frame 5). Scale bar is 300 μmdroplet. Frames 2–5 are repeated for the other concentrations of IL, which(frame 6). These droplets capture a single cell (from the generator) and(See ESI† Video S1)

As shown in Fig. 1, a key feature of our design is matingof two droplet devices via a capillary. Droplets with a singlecell are generated in the droplet-in-channel chip and arepressure-driven through the capillary and translated ontothe digital microfluidic device (Fig. 1d; see exploded viewin ESI† Fig. S1). The step-by-step process is shown in Fig. 2.Here, a droplet is generated in a biocompatible oil phase(i.e. HFE 7500) with 0.5% Picosurf surfactant (Fig. 2a). Weused this surfactant and oil as the combination has beenshown to be biocompatible (and we believe that the surfac-tant and oil aided our prevention for biofouling on the DMFsurface).16,22 The concentration of the surfactant was opti-mized to minimize merging of droplets inside the channel butat the same time allow merger with droplets containing thecell media and ionic liquid on the digital microfluidic device.As the droplet reaches the capillary inlet, it is transferred

This journal is © The Royal Society of Chemistry 2014

ages showing droplet encapsulation of a single cell. Fluorescent dyes from the generator are pressure-driven into the DMF device (frame 1).F electrodes (frames 3 and 4). The 20 nL droplet from the generator. (c) Frames 1–5 show the actuation scheme to generate a 150 mM ILresults in four droplets having a 0, 37.5, 75, or 150 mM IL concentrationare actuated to their own separate culture region. Scale bar is 2 mm.

Lab on a Chip Paper

Publ

ishe

d on

30

Oct

ober

201

4. D

ownl

oade

d by

Law

renc

e B

erke

ley

Nat

iona

l Lab

orat

ory

on 3

0/10

/201

4 13

:57:

29.

View Article Online

to the DMF device where it reaches a section containing ten(a 5 × 2 array) 300 μm × 300 μm inter-digitated electrodes.This droplet is immediately actuated and merged with aIL/salt droplet on the digital microfluidic device (shown inframes 1–5 in Fig. 2b). Serial dilutions of the IL/salt wereformed by merging a droplet (0.1 μL) containing cell mediaand a 300 mM IL/salt droplet. By mixing this droplet in acircular pattern,59 it produced a 2× dilution of the ionic liquid(frames 1–5 in Fig. 2c). This 150 mM droplet was then splitinto two droplets. One droplet was actuated to the cell culturezone and the other was used for subsequent mixing with adispensed cell media droplet resulting in droplets containing75 and 37.5 mM concentrations (frame 6 – Fig. 2c). Such oper-ations would be very difficult to employ on a microchanneldevice as it would require valving or injection/coalescencetechniques demanding optimal timing and control.

D2D automation system

An attractive feature of microfluidic platforms is easier auto-mation (via control of voltages and fluid flow) than conven-tional fluid-handling robots. Here, we designed and built anintegrated automated microfluidic system that is capable ofculturing cell in different conditions (Fig. 3). The platform

This journal is © The Royal Society of Chemistry 2014

Fig. 3 Automation system controlling the D2D microfluidic device. Thidroplet movement and sensing on the device, temperature control by theused to control the Peltier and to maintain constant temperature in the cuneMESYS syringe pumps to maintain steady flow rates and to prevent dro(http://www.codeblocks.org/home) using their default GCC compiler which

comprises four core components: a feedback system for drop-let sensing and detection, temperature control for culturingcells, flow rate control for droplet generation, and automateddroplet movement on the DMF device. Automating thesecomponents together is challenging, as sophisticated controlis required to ensure proper droplet movement and continu-ous cell culturing operation. Hence, a primary goal for thework reported here was the development of an automationsystem that is capable to automate single cell droplet genera-tion and movement and to culture cells with unattendedoperation for extended periods of time. Shih et al.41,47,48

recently reported an impedance sensing and feedback controlsystem for high-fidelity droplet movement and sensing ondigital microfluidic devices. The system is very useful for manip-ulating “sticky” droplets such as those containing proteins. Weadapt this system for droplet actuation and sensing dropletson the D2D microfluidic system and incorporate additionalfeatures (e.g. controlling flow rates) to ensure compatibilitywith other microfluidic paradigms (e.g. droplet-in-channels)and continuous cell culturing and analysis.

The automation workflow consists of executing five proce-dures: 1) flow rate control, 2) droplet sensing, 3) IL dilutions,4) mixing and cell culture, and 5) evaporation prevention.The flow control consists of controlling a high-precision

Lab Chip

s schematic shows the components and the signal flow that controlPeltier element, and flow rates from the syringe pump. The Arduino islture regions on the DMF. An in-house C++ program is used to controlplet evaporation. Scripts are written in C++ using Codeblocks 10.05is included as ESI.†

Lab on a ChipPaper

Publ

ishe

d on

30

Oct

ober

201

4. D

ownl

oade

d by

Law

renc

e B

erke

ley

Nat

iona

l Lab

orat

ory

on 3

0/10

/201

4 13

:57:

29.

View Article Online

neMESYS syringe pump and flow is initially applied to startgeneration of the droplets. Once the droplets are created andcontain a single cell, they are pressure driven to the inter-digitated 300 μm electrodes on the DMF device via capillary.During the droplet generation phase, the feedback systemcontinually (every 200 ms) applies 200 V (operating at 8 kHz)pulses to the 300 μm electrodes to check if a droplet fromthe generator is on the electrode. When a droplet is “sensed”on the electrode (i.e., detecting the difference in impedancewhen the electrode is not bearing a droplet and when thereis a droplet on the electrode), it immediately stops the flowand the droplet containing a single cell is mixed with a drop-let containing IL (IL dilutions were generated on the DMFprevious to the flow rate control – see Fig. 2c and ESI† video).These four droplets are actuated to the cell culture zone andthe Peltier heater was immediately ramped to its desired tem-perature (30 °C). To culture the cells on DMF, we started amixing sequence where we would circulate the droplet withcells around the 2 × 3 electrode array (incl. L-shaped electrode).Each actuation consists of 100 ms pulses and with a 500 msbreak step (i.e. no electrodes are actuated). In initial studies,we observed that without this break step, dielectric breakdownwould occur during the cycling of the droplet (~60% of thedevices; N = 15). However, with an added break step to themixing sequence, it eliminated the breakdown and we wereable to achieve high-fidelity droplet movement. To preventevaporation of the droplets, the device was initially floodedwith oil (~1–2 mL of HFE 7500). The automation systeminjects oil between the top and bottom layers of the devicethrough a capillary using the channels fabricated on the

Lab Chip

Fig. 4 A single S. cerevisiae cell growing in ionic liquid ([C2mim] [Cl]). Celwere maintained at 30 °C using an automated Peltier element for 24 h andSeries of images were collected at four intervals (0, 2, 6, and 24 h) to monit

DMF device during cell culture. To start the oil infusion pro-cess, an electric potential of 200 V at 8 kHz is applied to a ref-erence electrode (i.e. any electrode that does not hold adroplet) every 5 s. Two parameters are used for the infusionprocess: Voil and Vmeasured. Voil represents the measured volt-age across the reference electrode not bearing a droplet – inother words, with only oil between the top and bottom plates(we measured Voil = 0.174 ± 0.014 V; N = 12 on our devices).Vmeasured is the voltage of the reference electrode during operation.If Vmeasured < Voil (which represents the reference electrodebeing exposed to air – oil layer is evaporating), then oilis injected into the device at a flow rate of 10 μL s−1. Theflow would stop when the reference electrode reaches thethreshold value (i.e. Vmeasured > Voil). Two syringes were usedfor this process and were each equipped with a 2-way valvethat allows refill of the syringes and injection of oil into thedevice. Only one was used during operation and when itbecomes empty, the other filled syringe is activated while theempty syringe is set to refill. This new automation system wascapable of implementing flow control of droplets, temperaturecontrol, droplet manipulation and movement, and preventingdroplet evaporation. We speculate that this approach will beuseful for applications that require long incubation periodsand automation (e.g. cell culturing in different conditions,screening clinical samples, or enzymatic chemical reactions).

Impact of IL on growth and ethanol production

We used the automated D2D microfluidic device to deter-mine the effect of IL/salt type and concentration on the

This journal is © The Royal Society of Chemistry 2014

ls were grown at four concentrations of four types of IL/salt. The cellsthe droplet was circulated in the culture region to minimize clumping.or growth. Bar, 20 μm.

Lab on a Chip Paper

Publ

ishe

d on

30

Oct

ober

201

4. D

ownl

oade

d by

Law

renc

e B

erke

ley

Nat

iona

l Lab

orat

ory

on 3

0/10

/201

4 13

:57:

29.

View Article Online

morphological structure and growth of a single yeast cell andcompared it to conventional well-plate experiments. To con-duct a microfluidic screen, the cultures were interrogatedwith four concentrations of each IL or salt (the automatedscreening workflow is shown in ESI† Fig. S2). For both platforms,the culture conditions (e.g. Peltier heater temperature andIL concentration) were maintained the same and a total of16 IL/salt conditions were evaluated. To minimize clumpingof yeast, we circulated the droplet of yeast cells in a circularpattern (as described in the methods section) during culturing

This journal is © The Royal Society of Chemistry 2014

Fig. 5 Growth curves for yeast cultured in four types of ionic liquid/salt amicrofluidic device (right). Well plate growth curves were generated by aobtained every 10 min up to 24 h. Each measurement was evaluated in tr40× microscope images and yeast cells were counted manually or using Imerror bars represent one standard deviation.

on the D2D device using the automation system. The single cellproliferation images cultured in [C2mim] [Cl] are shown inFig. 4. The morphology of the yeast for [C2mim] [Cl], NaCl,and NaOAcs were unchanged during the 24 h at all tested con-centrations. [C2mim] [OAc] at 75 mM and at 150 mM concen-trations led to lysing of cells indicating that the ionic liquid istoxic to the yeast and will inhibit growth at these concentra-tions (see ESI† Fig. S3). Fig. 5 and Table S3† depict the growthcurves and the growth rates respectively for the IL conditionstested in the well-plate and in the D2D microfluidic device.

Lab Chip

nd four different concentrations in the well-plate (left) and in the D2Dwell plate reader (Tecan F200) and absorbance measurements were

iplicate. D2D microfluidic growth curves were generated by obtainingage J. Samples using D2D microfluidics were evaluated in triplicate and

Lab on a ChipPaper

Publ

ishe

d on

30

Oct

ober

201

4. D

ownl

oade

d by

Law

renc

e B

erke

ley

Nat

iona

l Lab

orat

ory

on 3

0/10

/201

4 13

:57:

29.

View Article Online

The growth curves shown in Fig. 5 are in qualitative agree-ment with respect to their growth profile (i.e. no IL showslargest growth at ~24 h) for both platforms. The replicates ofthe control experiments (i.e. intra-experiments) showed goodreproducibility whereas the inter-experiments (experimentsbetween different IL/salt) showed some variation at 2 h and6 h on the D2D device ( p = 0.004 and p = 0.001, respectively)but not in well-plates (see ESI† Fig. S4). We speculate the cellsare exposed to fluctuating temperatures during preparation,in the syringe pumps, and in the droplet generator. Thesefluctuating conditions can directly affect their growth rates asthe yeast travels through these various zones. Another causefor heterogeneity could be that the encapsulated single cellis at a different stage of its life cycle. In addition, growth waslow or inhibited at concentrations greater than 37.5 mM for[C2mim] [OAc] and at 150 mM for NaOAc. Some studies haveshown that acetic acid in its dissociated form is known tohave an inhibitory impact on microbial growth60 which couldexplain the differences in morphology at 0 h and at 24 h andthe decrease in their growth at these IL concentrations. Forthe other ionic liquids/salts tested, it showed no inhibitoryeffect for concentrations below 37.5 mM (where growth wassimilar in trend as the control).

We calculated the growth rates for yeast grown in bothwell-plates and D2D platforms and determined that somecalculated growth rates (in well-plates) were very similar invalues. For example, yeast cultured in NaOAc have growthrates of 0.228 h−1, 0.268 h−1, and 0.241 h−1 for 0, 37.5, and75 mM concentrations respectively; whereas for yeast cultured

Lab Chip

Fig. 6 Ethanol production of yeast cultured in a) NaOAc, b) NaCl, c) [Canaerobic conditions for additional 24 h. The inset graph in b) shows theobtained using an Agilent 1200 HPLC and were evaluated in triplicate. A cwith excellent linearity (R2 = 0.9999; standard curve not shown).

in D2D, we obtained 0.510 h−1, 0.444 h−1, and 0.379 h−1

for the same concentrations respectively. These disparitiesin the growth rates of the D2D device were of statisticaldifference ( p = 0.015) while the well-plate methodswere not statistically significant ( p = 0.62). Dewan et al.58 andWilson et al.61 recently reported that the heterogeneity ofpopulation of cells is less prominent compared to single cellsdue to “averaging out” the population heterogeneity58 or cell-to-cell variation in the single cell analysis.61 Therefore, wehypothesize that these effects may explain the differences ingrowth rates described here.

We also studied the impact of IL type and concentrationon ethanol production in yeast (see description in Fig. S2†).We evaluated the influence of IL/salt on ethanol production at16 conditions (same as above). Cultures were grown on theD2D microfluidic device for 24 h and then cultured in anaerobicchamber for another 24 h. Ethanol concentration was evaluatedby HPLC. Ethanol eluted at ~20 min and was well separatedfrom the other constituents (as shown in the chromatogramFig. S5†). The ethanol production for each ionic liquid at fourdifferent concentrations is shown in Fig. 6. The total ethanolproduction for each ionic liquid decreased with increasingIL concentration (Fig. 6a–d). At higher concentrations of ionicliquid (75 and 150 mM), NaCl, NaOAc, and [C2mim] [Cl]exhibited a slight decrease (1.2–2×) in their ethanol production,while ionic liquid containing 1-ethyl-3-methylimidazoliumacetate (i.e. [C2mim] [OAc]) showed a significant decrease(150 mM resulted in 58× decrease) in their ethanol produc-tion compared to the control (0 mM). We plotted the ethanol

This journal is © The Royal Society of Chemistry 2014

2mim] [OAc], and d) [C2mim] [Cl] on the D2D device for 24 h and inethanol production per cell when cultured in [C2mim] [OAc]. Data wasalibration curve was generated through the use of external standards

Lab on a Chip Paper

Publ

ishe

d on

30

Oct

ober

201

4. D

ownl

oade

d by

Law

renc

e B

erke

ley

Nat

iona

l Lab

orat

ory

on 3

0/10

/201

4 13

:57:

29.

View Article Online

produced per cell for this ionic liquid (Fig. 6c inset) andit shows at 75 and 150 mM only produces 0.03–0.04 g L−1

of ethanol per cell. This is much lower compared to the otherionic liquid and salts which produce 0.15–0.55 g L−1 of ethanolper cell at these high concentrations. This observation isexpected since [C2mim] [OAc] at these high concentrations isknown to be toxic and to affect the fermentative metabolism.62

Development of an engineered IL-tolerant host can potentiallyeliminate or reduce the adverse effects of these solvents.

Conclusion

We have developed a novel automated droplet-to-digital (D2D)microfluidic platform for screening the effect of ionic liquidson cell growth and ethanol production. We also developed anautomation system for conducting serial dilutions withoutpipetting, control temperature, and vary pressure flow (forpreventing droplet evaporation and for delivering dropletswith a single cell). In addition, this platform allows for theaddition of reagents in parallel and single cell analysis. Wetested our platform to screen multiple IL conditions thatshowed [C2mim] [OAc] had inhibitory effects on yeast growthand therefore significantly reduced their ethanol production.We also showed from statistical analysis that conductingsingle cell analysis using microfluidics resulted in observingdifferences in growth kinetics, which was not pronounced inthe well-plate analysis. Here we demonstrate a 4-plex formatbut much higher level of multiplexing can be achieved usinghigher number of pixels in the array63 or using bussed electrodeconfigurations.64 Furthermore, we have shown that this methodallows a 600-fold reduction in volume compared to well-plateformats (which require from 200–800 μL) while our assay onlyrequires a total of ~120 nL. Although the single cell screeningdemonstrated here is focused on studying ionic liquid withbiofuel applications, the automated microfluidic device demon-strated are applicable to many different organisms and may proveuseful for a diverse range of applications in single cell analysesthat require frequent reagent addition and mixing steps.

Acknowledgements

The authors thank Marijke Frederix and Dr. AindrilaMukhopadhyay for the yeast isolate; Susan Yilmaz for initialhelp with microscopy; Jennifer Gin, Edward Baidoo, Helcio Burdfor their assistance with operating the HPLC and chromatogramanalysis; This work was part of the DOE Joint BioEnergy Institute(www.jbei.org), supported by the US Department of Energy, Officeof Science, Office of Biological and Environmental Research,through contract DE-AC02-05CH11231 between LawrenceBerkeley National Laboratory and the US Department of Energy.Sandia is a multi-program laboratory operated by SandiaCorporation, a Lockheed Martin Company, for the United StatesDepartment of Energy's Nuclear Security Administration undercontract DE-AC04-94AL85000.

This journal is © The Royal Society of Chemistry 2014

References

1 B. A. Simmons, D. Loque and H. W. Blanch, Genome Biol.,

2008, 9, 242.

2 P. P. Peralta-Yahya and J. D. Keasling, Biotechnol. J., 2010, 5,

147–162.

3 A. P. Dadi, S. Varanasi and C. A. Schall, Biotechnol. Bioeng.,

2006, 95, 904–910.

4 R. P. Swatloski, S. K. Spear, J. D. Holbrey and R. D. Rogers,

J. Am. Chem. Soc., 2002, 124, 4974–4975.

5 S. Singh, B. A. Simmons and K. P. Vogel, Biotechnol. Bioeng.,

2009, 104, 68–75.

6 C. Li, B. Knierim, C. Manisseri, R. Arora, H. V. Scheller,

M. Auer, K. P. Vogel, B. A. Simmons and S. Singh, Bioresour.Technol., 2010, 101, 4900–4906.

7 D. Klein-Marcuschamer, P. Oleskowicz-Popiel, B. A. Simmons

and H. W. Blanch, Biomass Bioenergy, 2010, 34, 1914–1921.

8 D. Klein-Marcuschamer, P. Oleskowicz-Popiel, B. A. Simmons

and H. W. Blanch, Biotechnol. Bioeng., 2012, 109, 1083–1087.

9 Q. Li, Y. C. He, M. Xian, G. Jun, X. Xu, J. M. Yang and

L. Z. Li, Bioresour. Technol., 2009, 100, 3570–3575.

10 M. Zavrel, D. Bross, M. Funke, J. Buchs and A. C. Spiess,

Bioresour. Technol., 2009, 100, 2580–2587.

11 H. Pfruender, R. Jones and D. Weuster-Botz, J. Biotechnol.,

2006, 124, 182–190.

12 Q. Huang, Q. Wang, Z. Gong, G. Jin, H. Shen, S. Xiao,

H. Xie, S. Ye, J. Wang and Z. K. Zhao, Bioresour. Technol.,2013, 130, 339–344.

13 G. B. Fogel and C. F. Brunk, Anal. Biochem., 1998, 260, 80–84.

14 W. A. Duetz, L. Ruedi, R. Hermann, K. O'Connor, J. Buchs

and B. Witholt, Appl. Environ. Microbiol., 2000, 66, 2641–2646.15 T. Schneider, J. Kreutz and D. T. Chiu, Anal. Chem., 2013, 85,

3476–3482.16 L. Mazutis, J. Gilbert, W. L. Ung, D. A. Weitz, A. D. Griffiths

and J. A. Heyman, Nat. Protoc., 2013, 8, 870–891.17 T. Kong, J. Wu, K. W. Yeung, M. K. To, H. C. Shum and

L. Wang, Biomicrofluidics, 2013, 7, 44128.18 Z. Liu and H. C. Shum, Biomicrofluidics, 2013, 7, 44117.

19 D. Saeki, S. Sugiura, T. Kanamori, S. Sato and S. Ichikawa,

Lab Chip, 2010, 10, 357–362.20 S. Koster, F. E. Angile, H. Duan, J. J. Agresti, A. Wintner,

C. Schmitz, A. C. Rowat, C. A. Merten, D. Pisignano, A. D. Griffithsand D. A. Weitz, Lab Chip, 2008, 8, 1110–1115.

21 J. Clausell-Tormos, D. Lieber, J. C. Baret, A. El-Harrak,

O. J. Miller, L. Frenz, J. Blouwolff, K. J. Humphry, S. Koster,H. Duan, C. Holtze, D. A. Weitz, A. D. Griffiths and C. A. Merten,Chem. Biol., 2008, 15, 427–437.

22 E. Brouzes, M. Medkova, N. Savenelli, D. Marran,

M. Twardowski, J. B. Hutchison, J. M. Rothberg, D. R. Link,N. Perrimon and M. L. Samuels, Proc. Natl. Acad. Sci. U. S. A.,2009, 106, 14195–14200.

23 C. Chang, J. Sustarich, R. Bharadwaj, A. Chandrasekaran,

P. D. Adams and A. K. Singh, Lab Chip, 2013, 13, 1817–1822.

24 L. Mazutis, J. C. Baret, P. Treacy, Y. Skhiri, A. F. Araghi,

M. Ryckelynck, V. Taly and A. D. Griffiths, Lab Chip, 2009, 9,2902–2908.

Lab Chip

Lab on a ChipPaper

Publ

ishe

d on

30

Oct

ober

201

4. D

ownl

oade

d by

Law

renc

e B

erke

ley

Nat

iona

l Lab

orat

ory

on 3

0/10

/201

4 13

:57:

29.

View Article Online

25 S. L. Sjostrom, Y. Bai, M. Huang, Z. Liu, J. Nielsen, H. N. Joensson

and H. Andersson Svahn, Lab Chip, 2014, 14, 806–813.

26 L. Yu, M. C. Chen and K. C. Cheung, Lab Chip, 2010, 10,

2424–2432.

27 D. Pekin, Y. Skhiri, J. C. Baret, D. Le Corre, L. Mazutis,

C. B. Salem, F. Millot, A. El Harrak, J. B. Hutchison,J. W. Larson, D. R. Link, P. Laurent-Puig, A. D. Griffiths andV. Taly, Lab Chip, 2011, 11, 2156–2166.

28 L. Mazutis, J. C. Baret and A. D. Griffiths, Lab Chip, 2009, 9,

2665–2672.

29 H. Song, M. R. Bringer, J. D. Tice, C. J. Gerdts and

R. F. Ismagilov, Appl. Phys. Lett., 2003, 83, 4664–4666.

30 B. Ahn, K. Lee, R. Panchapakesan and K. W. Oh, Biomicrof luidics,

2011, 5, 24113.

31 B. O'Donovan, D. J. Eastburn and A. R. Abate, Lab Chip,

2012, 12, 4029–4032.

32 D. J. Eastburn, A. Sciambi and A. R. Abate, PLoS One, 2013,

8, e62961.

33 L. Xu, H. Lee, R. Panchapakesan and K. W. Oh, Lab Chip,

2012, 12, 3936–3942.

34 X. Niu, S. Gulati, J. B. Edel and A. J. deMello, Lab Chip, 2008,

8, 1837–1841.

35 A. R. Wheeler, Science, 2008, 322, 539–540.

36 K. Choi, A. H. Ng, R. Fobel and A. R. Wheeler, Annu. Rev.

Anal. Chem., 2012, 5, 413–440.37 N. Vergauwe, D. Witters, F. Ceyssens, S. Vermeir,

B. Verbruggen, R. Puers and J. Lammertyn, J. Micromech.Microeng., 2011, 21, 054026.

38 D. Witters, N. Vergauwe, S. Vermeir, F. Ceyssens, S. Liekens,

R. Puers and J. Lammertyn, Lab Chip, 2011, 11, 2790–2794.

39 P. T. Kumar, F. Toffalini, D. Witters, S. Vermeir, F. Rolland,

M. Hertog, B. M. Nicolai, R. Puers, A. Geeraerd andJ. Lammertyn, Sens. Actuators, B, 2014, 199, 479–487.

40 S. H. Au, S. C. C. Shih and A. R. Wheeler, Biomed. Microdevices,

2011, 13, 41–50.

41 S. C. C. Shih, I. Barbulovic-Nad, X. Yang, R. Fobel and

A. R. Wheeler, Biosens. Bioelectron., 2013, 42, 314–320.

42 I. Barbulovic-Nad, H. Yang, P. S. Park and A. R. Wheeler,

Lab Chip, 2008, 8, 519–526.

43 V. N. Luk, G. Mo and A. R. Wheeler, Langmuir, 2008, 24,

6382–6389.

44 S. H. Au, P. Kumar and A. R. Wheeler, Langmuir, 2011, 27,

8586–8594.

Lab Chip

45 S. H. Au, R. Fobel, S. P. Desai, J. Voldman and A. R. Wheeler,

Integr. Biol., 2013, 5, 1014–1025.

46 R. Rodríguez-Díaz, T. Wehr and S. Tuck, Analytical techniques

for biopharmaceutical development, M, Dekker, Boca Raton,2005.

47 S. C. C. Shih, R. Fobel, P. Kumar and A. R. Wheeler, Lab Chip,

2011, 11, 535–540.

48 S. C. C. Shih, H. Yang, M. J. Jebrail, R. Fobel, N. McIntosh,

O. Y. Al-Dirbashi, P. Chakraborty and A. R. Wheeler, Anal. Chem.,2012, 84, 3731–3738.

49 M. Abdelgawad, M. W. Watson and A. R. Wheeler, Lab Chip,

2009, 9, 1046–1051.

50 M. W. Watson, M. J. Jebrail and A. R. Wheeler, Anal. Chem.,

2010, 82, 6680–6686.

51 J. Gorbatsova, M. Jaanus and M. Kaljurand, Anal. Chem.,

2009, 81, 8590–8595.

52 H. Kim, M. S. Bartsch, R. F. Renzi, J. He, J. L. Van de Vreugde,

M. R. Claudnic and K. D. Patel, J. Lab. Autom., 2011, 16, 405–414.

53 H. Kim, M. J. Jebrail, A. Sinha, Z. W. Bent, O. D. Solberg,

K. P. Williams, S. A. Langevin, R. F. Renzi, J. L. Van De Vreugde,R. J. Meagher, J. S. Schoeniger, T. W. Lane, S. S. Branda,M. S. Bartsch and K. D. Patel, PLoS One, 2013, 8, e68988.

54 N. Thaitrong, H. Kim, R. F. Renzi, M. S. Bartsch, R. J. Meagher

and K. D. Patel, Electrophoresis, 2012, 33, 3506–3513.

55 R. B. Fair, Microfluid. Nanofluid., 2007, 3, 245–281.

56 G. H. Markx, Organogenesis, 2008, 4, 11–17. 57 B. L. Wang, A. Ghaderi, H. Zhou, J. Agresti, D. A. Weitz,

G. R. Fink and G. Stephanopoulos, Nat. Biotechnol., 2014, 32,473–478.

58 A. Dewan, J. Kim, R. H. McLean, S. A. Vanapalli and M. N. Karim,

Biotechnol. Bioeng., 2012, 109, 2987–2996.

59 P. Paik, V. K. Pamula and R. B. Fair, Lab Chip, 2003, 3, 253–259.

60 B. Maiorella, H. W. Blanch and C. R. Wilke, Biotechnol. Bioeng.,

1983, 25, 103–121.61 J. L. Wilson, S. Suri, A. Singh, C. A. Rivet, H. Lu and

T. C. McDevitt, Biomed. Microdevices, 2014, 16, 79–90.62 Y. V. Nancharaiah and A. J. Francis, Bioresour. Technol.,

2011, 102, 6573–6578.63 B. Hadwen, G. R. Broder, D. Morganti, A. Jacobs, C. Brown,

J. R. Hector, Y. Kubota and H. Morgan, Lab Chip, 2012, 12,3305–3313.

64 S. C. C. Shih, N. S. Mufti, M. D. Chamberlain, J. Kim and

A. R. Wheeler, Energy Environ. Sci., 2014, 7, 2366–2375.

This journal is © The Royal Society of Chemistry 2014